Abstract
Chronic pain is exacerbated by maladaptive cognition such as pain catastrophizing (PC). Biomarkers of PC mechanisms may aid precision medicine for chronic pain. Here, we investigate EEG biomarkers using mass univariate and multivariate (machine learning) approaches. We test theoretical notions that PC results from a combination of augmented aversive-value encoding ("magnification") and persistent expectations of pain ("rumination"). Healthy individuals with high or low levels of PC underwent an experimental pain model involving nociceptive laser stimuli preceded by cues predicting forthcoming pain intensity. Analysis of EEG acquired during the cue and laser stimulation provided event-related potentials (ERPs) identifying spatially and temporally-extended neural representations associated with pain catastrophizing. Specifically, differential neural responses to cues predicting high vs. low intensity pain (i.e. aversive value encoding) were larger in the high PC group, largely originating from mid-cingulate and superior parietal cortex. Multivariate spatiotemporal EEG patterns evoked from cues with high aversive value selectively and significantly differentiated the high PC from low PC group (64.6% classification accuracy). Regression analyses revealed that neural patterns classifying groups could be partially predicted (R-2 = 28%) from those neural patterns classifying the aversive value of cues. In contrast, behavioural and EEG analyses did not provide evidence that PC modifies more persistent effects of prior expectation on pain perception and nociceptive responses. These findings support the hypothesis of magnification of aversive value encoding but not persistent expression of expectation in pain catastrophizers. Multivariate patterns of aversive value encoding provide promising biomarkers of maladaptive cognitive responses to chronic pain that have future potential for psychological treatment development and clinical stratification.